FREE TOOLKIT
Nine self-check tools for industrial technology decisions, built on Thing Company's published papers. They run inside the AI assistant you already use, including Claude, ChatGPT and Codex, and grade your evidence the way a Findings Brief would: Verified, Benchmarked or Assumption.
The first tool, Start here, asks which moment your decision is in and who you are, then points you to the tools below. The rest follow the same three moments as this site.
Any moment
Asks which moment your decision is in and whether you are a manufacturer, an equipment maker, a technology company or an investor, then names the two or three tools to run first.
Any buyer
Runs your bet through whose job changes, who funds it and how finance books it, who else can say no, who will run it, and why it wins the budget this year, with each answer graded.
Any buyer
Grades each line of your business case Verified, Benchmarked or Assumption, and names the risk behind each Assumption and a way to test it.
Any buyer
Builds a guide in five sections that asks about the last real purchase rather than a hypothetical one, including the budget, the approval path and the decision date.
For investors
Turns an investment thesis into the four to six claims the return depends on, writes each as a pass/fail criterion, and states what a diligence window can evidence.
Any buyer
For a pilot that worked and has not scaled: separates a wrong hypothesis from weak execution, so you know whether to fix the plan, fix the delivery, or stop.
For operators
One page per plant before a network rollout: workforce adoption, maintenance and support, controls, network and security, and the value case each plant controller will sign.
For technology companies
Tests a partner plan against the claims inside it: the partner holds the relationship, can implement, earns enough, avoids conflict and stays accountable.
For OEMs
Rehearses the first payment dispute on an uptime guarantee or outcome contract before you sign it, using one month of data that both sides calculate.
THE TOOLKIT
All nine tools in one download, plus each tool as its own file for Claude. Tell us who you are and where your decision stands, and the files are yours.
We use these details to understand who uses the toolkit and, only with your consent, to send follow-up email. See the privacy policy .
Claude, on the web and desktop. Add one tool at a time. Open Customize, then Skills, select the plus button, choose Create skill, then Upload a skill, and upload a tool's ZIP file. Each ZIP holds one skill folder named for the skill. Skills are available on every plan, with code execution enabled.
Claude Code and Cowork. Install the plugin file, thing-company-toolkit.plugin. All nine tools load together.
ChatGPT desktop and Codex. Unzip the skills folder into ~/.agents/skills, or into a project's .agents/skills folder. In Codex, type /skills or $ to call a tool. Standalone skills work in the ChatGPT desktop app, the Codex CLI and the Codex IDE extension.
Cursor, GitHub Copilot in VS Code, Gemini CLI and other tools. These support the open Agent Skills standard. Unzip the skills folder where each tool's documentation says: Cursor, GitHub Copilot in VS Code, Gemini CLI. The standard and the tools that support it are listed at agentskills.io.
Any other assistant. Open a tool's SKILL.md file, paste its text into the chat, and ask the assistant to follow it.
Start a tool. In Claude and Claude Code, ask in plain words, such as "help me check our rollout before the capital request", or name the tool: "run the five buyer questions check". In Codex, type /skills or $ and pick the tool; in the ChatGPT desktop app, open Skills in the sidebar. In any other assistant, paste the tool's SKILL.md text into the chat and say "run this".
Not sure which one? Run Start here first. It asks which moment your decision is in and who you are, then names the two or three tools to run.
Example openings, one per tool. Replace the brackets with your own details.
What you get back. A working document, not a chat: a graded gap list, an interview guide, a site sheet, a set of pass/fail criteria or a dispute log. Every claim is graded Verified, Benchmarked or Assumption, and each Assumption comes with the next test.
An illustrative example. A plant operations lead runs the five buyer questions check on a vision inspection rollout. Two answers come back Verified: the plant managers' workaround cost and the operating budget for year one. Two are Benchmarked: payback from a vendor case and integrator coverage from a partner's map. One is an Assumption: nobody has asked OT security at the other plants. The next step the tool names is that conversation. This example is illustrative, not a client engagement.
Tips. State the bet in one sentence before you start. Use placeholders for customer, site and partner names. Nothing is stored, so save the conversation if you want to keep the result.
Bring the result to a Sprint Assessment. The gaps a tool finds are the claims a Sprint tests with the buyers who would sign. → Request a Sprint Assessment
Each tool grades your own evidence the way a Findings Brief would. It cannot test the buyers who would sign, and it cannot be independent of you: the answers it grades are the ones you give it.
The tools are text instructions that run inside your conversation with your assistant. They send nothing to Thing Company and store nothing; your assistant's own data terms apply. The form above is the only thing we receive.
Yes. We ask for your name, work email, company, role, the moment your decision is in and when it has to be made, so we know who uses it. Apart from the email with your download links, we send follow-up only if you tick the box.
Start here. It asks which moment your decision is in and who you are, then names the two or three tools to run first. If you already know your moment: the five buyer questions check before a commitment, the stall diagnosis for a pilot that has not scaled, and the rollout site sheet before a network rollout.
Not for Claude: skills are available on every Claude plan, with code execution enabled. In ChatGPT, standalone skills work in the desktop app, the Codex CLI and the Codex IDE extension. In any assistant, you can paste a tool's SKILL.md text into the chat instead.
Keep it. It is a working document you can share with your team. If the gaps it finds sit under a decision you have to take, a Sprint Assessment is thirty minutes on whether they are worth testing with the buyers who would sign.
Claude on the web and desktop, Claude Code and Cowork, the ChatGPT desktop app and Codex, and other tools that support the open Agent Skills standard, such as Cursor, GitHub Copilot in VS Code and Gemini CLI. Any other assistant can follow a tool if you paste its SKILL.md text into the chat.
No. The tools are text instructions that run inside your assistant, and nothing you type into them reaches Thing Company. Your assistant's own data terms apply.
No. It grades the evidence you already have. A Sprint tests the claims independently with the buyers who would sign and ends in a recommended verdict, which you then decide on.
Twenty-four hours. Submit the form on this page again for fresh links.
A self-check shows where your evidence is thin. A Sprint Assessment is thirty minutes on whether it is worth testing with the buyers who would sign.
Or reach us directly at hello@thing.company